Diagnostics of convergence of an estimate of the marginal density via calibrated divergence measures

Abstract

This paper presents a diagnostic procedure based on calibrated divergence measures to study the performance of a Rao‐Blackwellized estimate (RB) of the marginal density. The Gibbs sampling approach is used to compute the divergence measures. A new formulation of the Anchored Ratio Convergence Criterion (ARC ²) and the Difference Convergence Criterion (DC ²) introduced by Zellner and Min (1995) is considered in terms of the calibrated divergence measures. Some examples are provided to illustrate these procedures.

Keywords:

• Computational statistics
• Gibbs sampler
• ⊘‐divergence